AMD
Operating mode
Use this to bias workload recommendations toward responsiveness, background autonomy, lighter serving, or multi-GPU scale-out.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
AMD Instinct MI325X 256GB 是 AMD 增强版基于 CDNA 4 的数据中心 GPU，提供 256 GB HBM3e 内存和 6 TB/s 带宽，在 MI300X 架构基础上增加了内存容量和带宽，面向最大规模的生产 LLM 推理工作负载。1307 TFLOPS FP16 算力配合更大内存包络，可以全精度运行最大的开源模型（405B、Llama-3-405B），无需量化。
Beyond LLMs
What AI tasks this GPU can handle — from text generation to image and video creation.
| Capability | Status | Representative Model |
|---|---|---|
| LLM Chat (7B) | Runs natively | Llama 3.1 8B Q4 |
| LLM Coding (30B) | Runs natively | Qwen 3 30B Q4 |
| LLM Large (70B) | Runs natively | Llama 3.1 70B Q4 |
| Image Gen (SDXL) | Runs natively | SDXL 1.0 FP16 |
| Image Gen (Flux) | Runs natively | Flux.1 Dev FP16 |
| Image Gen (SD 3.5) | Runs natively | SD 3.5 Large FP16 |
| Video Short (25f) | Runs natively | LTX Video 2B |
| Video Long (100f) | Runs natively | Wan Video 14B |
Architecture
CDNA 4 powers the next-generation Instinct MI325X and MI350X accelerators. Built on TSMC 3nm with up to 288 GB HBM3e memory and native FP4 support for maximum inference density.
AI Relevance
With up to 288 GB HBM3e and FP4 support, CDNA 4 targets the highest-density AI inference deployments. Directly competes with NVIDIA Blackwell B200 for large-scale model serving.
购买建议
本地 AI 的绝佳选择
能良好运行 50 个顶级模型中的 41 个 — 本地推理的全能之选。
256.0 GB
VRAM
$20,000
建议零售价
$78/GB
每 GB VRAM 成本
最适合此 GPU 的模型
What will limit you first
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
Best upgrade itinerary
Unlocks 1 additional models that do not fit on the current setup.
想要更多余量？ AMD Instinct MI350X 288GB (288.0 GB VRAM) 是下一步升级选择。
Chat
SThis model is a direct match for chat. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
Coding
SThis model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface.
Agentic Coding
SThis model is still usable for agentic-coding, but it is not the most specialized pick. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface.
Reasoning
SThis model is a direct match for reasoning. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface.
RAG
SThis model is a direct match for rag. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, lm-studio.
DeepSeek V4 Flash
Qwen 3 235B A22BQwen 3.5 122B A10B
Devstral 2 123B InstructMistral Small 4 119B
GPT-OSS 120B
Command A 111BDeepSeek Coder V2 236BPixtral Large 124BQwen3-Coder 30B A3B InstructQwen 2.5 VL 72BQwen 3.6 35B A3BQwen 3.5 27BQwen3-Coder-NextQwen 3.6 27BQwen3-VL 30B A3B InstructQwen 3.5 35B A3BQwen 3 32BMagistral Small 2507Devstral Small 2 24B InstructLeanstral 119B A6BQwen 3.5 9BQwen 3 30B A3B
Nemotron 3 Nano 30BQwen 3 14BDevstral Small 1.1
Phi-4-reasoning-plus 14BQwen 3 8BQwen 3.5 397B A17BGPT-OSS 20B
Gemma 4 31BNemotron Cascade 2 30B A3BQwen 3.5 4B
EXAONE 4.0 32BGemma 4 26B A4BMinistral 3 14BPhi-4 Mini Reasoning 4BNemotron Nano 8B
Jina Embeddings v3
BGE M3
Kimi K2.5Kimi K2.6DeepSeek V4 Pro
GLM-5.1GLM-5DeepSeek V3.2Qwen3-Coder 480B A35B InstructDeepSeek R1 671BDeepSeek V3.1 671B触手可及
高质量模型，只需稍多一点内存
Image & Video Generation
52 of 52 models can generate images or video on your AMD Instinct MI325X 256GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | 0ms | S |
| Stable Diffusion 1.5Image | 512×768 | 100ms | S |
| Realistic Vision v5.1Image | 512×768 | 100ms | S |
| DreamShaper 8Image | 512×768 | 100ms | S |
| LCM DreamShaper v7Image | 512×768 | 0ms | S |
| PixArt-SigmaImage | 1024×1024 | 300ms | S |
| FramePack I2VVideo | 1280×720 | 500ms/frame | S |
| SDXL TurboImage | 512×512 | 0ms | S |
| SDXL LightningImage | 1024×1024 | 100ms | S |
| Stable Diffusion XL 1.0Image | 1024×1024 | 300ms | S |
| Playground v2.5Image | 1024×1024 | 400ms | S |
| RealVisXL v5.0Image | 1024×1024 | 300ms | S |
| DreamShaper XLImage | 1024×1024 | 300ms | S |
| Juggernaut XL v9Image | 1024×1024 | 300ms | S |
| Animagine XL 3.1Image | 1024×1024 | 300ms | S |
| Pony Diffusion V6 XLImage | 1024×1024 | 300ms | S |
| Animagine XL 4.0Image | 1024×1024 | 300ms | S |
| Illustrious XLImage | 1024×1024 | 300ms | S |
| Wan Video 2.1 1.3BVideo | 480×832 | 200ms/frame | S |
| Stable Diffusion 3.5 MediumImage | 1024×1024 | 500ms | S |
| Flux.2 Klein 4BImage | 1024×1024 | 100ms | S |
| LTX Video 2BVideo | 1280×720 | 200ms/frame | S |
| KolorsImage | 1024×1024 | 500ms | S |
| Stable CascadeImage | 1024×1024 | 700ms | S |
| AuraFlow v0.3Image | 1536×1536 | ~1.2s | S |
| Stable Diffusion 3.5 LargeImage | 1024×1024 | ~1.4s | S |
| Stable Diffusion 3.5 Large TurboImage | 1024×1024 | 300ms | S |
| CogVideoX 2BVideo | 720×480 | 200ms/frame | S |
| HunyuanVideoVideo | 720×1280 | 500ms/frame | S |
| ChromaImage | 1024×1024 | 300ms | S |
| Z-Image TurboImage | 1536×1536 | 300ms | S |
| Flux.1 DevImage | 1024×1024 | ~1.2s | S |
| Flux.1 SchnellImage | 1024×1024 | 200ms | S |
| LTX Video 13BVideo | 1280×720 | 500ms/frame | S |
| Flux.1 Kontext DevImage | 1024×1024 | ~1.3s | S |
| AnimateDiff v1.5.3Video | 512×768 | 100ms/frame | S |
| Cosmos Diffusion 7BVideo | 1024×576 | 400ms/frame | S |
| CogVideoX 5BVideo | 720×480 | 300ms/frame | S |
| Wan2.2 TI2V 5BVideo | 832×480 | 300ms/frame | S |
| Flux.2 Klein 9BImage | 1024×1024 | 100ms | S |
| Flux.1 Fill DevImage | 1024×1024 | ~1.1s | S |
| Mochi 1 PreviewVideo | 848×480 | 400ms/frame | S |
| HunyuanVideo 1.5Video | 720×1280 | 400ms/frame | S |
| Helios 14BVideo | 1280×720 | 500ms/frame | S |
| SkyReels V2 14BVideo | 1280×720 | 500ms/frame | S |
| Wan Video 2.1 14BVideo | 720×1280 | 500ms/frame | S |
| Wan Video 2.2 14BVideo | 720×1280 | 500ms/frame | S |
| Qwen ImageImage | 1024×1024 | 400ms | S |
| Qwen Image EditImage | 1024×1024 | 400ms | S |
| Flux.2 DevImage | 1024×1024 | ~12.4s | S |
| MAGI-1Video | 1280×720 | 600ms/frame | S |
| HunyuanImage 3.0Image | 1024×1024 | 800ms | S |
Image models estimated at 1024×1024 (28 steps, FP16). Video models estimated at 768×512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Multi-GPU scaling
Scale out with multiple GPUs for larger models. Infinity Fabric provides 896 GB/s inter-GPU bandwidth with 12% overhead.
| Config | Effective memory | Models that fit | Est. bandwidth |
|---|---|---|---|
| 1× AMD | 256 GB | 363/374 | 6,000 GB/s |
| 2× AMD | 512 GB | 371/374 | 10,560 GB/s |
| 4× AMD | 1024 GB | 374/374 | 21,120 GB/s |
| 8× AMD | 2048 GB | 374/374 | 42,240 GB/s |
Model counts use default quantization at coding workload settings. Multi-GPU scaling factor: 0.88× per additional GPU.
Upgrade paths
See what you unlock with more powerful hardware
升级选项
Unlocks 11 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 65%.
Infinity Fabric gives this scale-out path a cleaner inter-GPU story than PCIe-only builds.
~$20,000 MSRP
Unlocks 1 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 12%.
~$8,000 MSRP
AMD Instinct MI325X 256GB (256 GB VRAM) can run these top models: DeepSeek V4 Flash (score: 99/100), Qwen 3 235B A22B (score: 95/100), Qwen 3.5 122B A10B (score: 93/100). See the full compatibility list above.
AMD Instinct MI325X 256GB has 256 GB of VRAM available for AI model inference. This determines which models and quantization levels you can run locally.
Yes, AMD Instinct MI325X 256GB is excellent for running LLMs locally with top compatibility scores above 80/100.
For coding on AMD Instinct MI325X 256GB, we recommend DeepSeek V4 Flash. It achieves 94.4 tokens per second with 872K context window. This model is a direct match for coding. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface.
There are 2 upgrade path(s) from AMD Instinct MI325X 256GB: AMD Instinct MI325X 256GB, AMD Instinct MI350X 288GB. Upgrading would unlock larger models and faster inference speeds.
Yes, AMD Instinct MI325X 256GB with 256 GB of usable memory can run Flux.1 Dev at FP16 natively. Flux is a 12B parameter diffusion transformer that produces high-quality images. You can also run the Schnell variant for faster generation.
AMD Instinct MI325X 256GB (256 GB VRAM) can handle various AI generation tasks beyond LLMs. For image generation, SDXL and Stable Diffusion 3.5 run well. Flux.1 Dev also runs natively for state-of-the-art image quality. For video, LTX Video 2.3 can generate short clips. Check the AI Capability Matrix above for detailed compatibility.
AMD Instinct MI325X 256GB is excellent for AI image generation. With 256 GB of usable memory, it runs all major diffusion models including Flux.1, SDXL, and Stable Diffusion 3.5 at full precision. You can generate high-resolution images quickly and even handle video generation models.
Yes, AMD Instinct MI325X 256GB with 256 GB of usable memory can run Qwen 3.5 27B at Q8 (near-lossless, ~28.9 GB) or even FP16 (~55.4 GB) depending on your context needs. This setup provides an excellent experience with this model. Use Ollama or vLLM for best results.
With 256 GB VRAM on AMD Instinct MI325X 256GB, use Q8_0 for most models — it is near-lossless and you have the memory for it. For 70B+ models, Q6_K offers excellent quality. Reserve Q4_K_M for 100B+ models or when you need maximum context length.
AMD Instinct MI325X 256GB already has strong memory bandwidth, so the next limit is often memory capacity and context headroom rather than raw decode speed. For local LLMs, fit first and bandwidth second is the right mental model.
AMD Instinct MI325X 256GB supports up to 8× GPU scaling via Infinity Fabric at 896 GB/s. With 8× GPUs, you get 2048 GB effective memory with a 0.88× scaling factor per GPU. This enables running models like Kimi K2.5 and Kimi K2.6 that don't fit on a single card.
Infinity Fabric is recommended for AMD Instinct MI325X 256GB multi-GPU inference, providing 896 GB/s interconnect bandwidth with only 12% scaling overhead. PCIe-only setups work but have higher overhead (~25%) due to limited inter-GPU bandwidth.
Compare with similar
